Salesforce wants to “democratize” the development of artificial intelligence (AI) applications. That is, making AI available to as many business users as possible.
Consequently, it’s baked an AI system called Einstein into the Salesforce platform – such that its capabilities are available across its various clouds, including Sales Cloud, Service Cloud and Marketing Cloud.
“AI is out of reach for the vast majority of companies, because it’s really hard,” said John Ball, SVP and GM of Einstein, during a press conference revealing the initiative last week.
AI processes involve collecting data, siphoning it into machine-learning algorithms that data scientists must build and maintain. Then customers must have infrastructure to scale its AI applications.
“The last mile, where companies fall down, is you have to surface these insights and predictions and recommendations in the context of the business user,” Ball said.
Basically, AI is too hard for most companies to apply, which is why Salesforce hopes to “democratize” it.
From a practical standpoint, that means letting business users of widely varying skill levels harness Salesforce’s AI platform to build custom apps.
People who don’t know anything about programming can slap together various assignment or workflow rules to predict outcomes, recommend actions or automate certain activities. More technical developers – who might not know anything about deep learning – can train Einstein to recognize consumer sentiment patterns or images. And data scientists can use Einstein to build custom algorithms and put them into production, without having to worry about scale issues.
So how did Salesforce get this capability? By putting together a whole bunch of acquisitions including MetaMind (deep learning), PredictionIO (machine learning), EdgeSpring (analytics), Beyondcore (data discovery) and others.
But much of the fuel for Salesforce Einstein comes from the data it's collected over the years.
“We have that great data asset where we’re collecting millions of signals from users and transactions every day,” said Salesforce Chief Scientist Richard Socher (formerly CEO and founder of MetaMind). “That rich data is driving our AI.”
There’s also an element to which Einstein has some functions of a data management platform (DMP) – a key piece of technology that Salesforce lacks, despite investing heavily in its Marketing Cloud.
Socher described an Einstein Marketing Cloud possibility: It could determine how web signals should influence messaging in channels like email, mobile or social.
“These scores segment an audience,” Socher said. “They segment an audience around what is least likely and most likely to convert, and least likely and most likely to open an email.”
(Marketing Cloud honcho Bob Stutz had a few more scenarios in a blog post.)
It’s still unclear how Einstein’s data processing and segmentation capabilities compare to a dedicated DMP like Krux, for instance – or whether incorporating a standalone DMP might make Einstein more powerful still.
Regardless, AI has made a big comeback in the marketing/ad tech community. Besides Salesforce, IBM emphasizes its Watson capability, which it merged with its Weather Co. assets. As with Salesforce’s Einstein, IBM’s Watson relies on historical data as a key component to make predictions and recommendations.
And Oracle also got into AI on Monday at its Open World conference with its “Adaptive Intelligent Applications," designed to use Oracle data to learn about consumers and target messaging appropriately.
And it’s not just the big guys: A host of smaller marketing tech companies like Adgorithms, Boomtrain and Cognitiv offer AI products as well.